[Tagging] tree rows vs individual trees

[Ture Pålsson]

2019-02-09 15:23 skrev Tom Pfeifer:

> If a renderer wants to cluster any trees that can be done 
> algorithmically.

As someone who tries to render smallish-scale (typcally 1:25000 or 
1:50000) maps from OSM data, I am always slightly annoyed when someone 
states that something does not need to be mapped bacuse it can be 
inferred algorithmically from other data, without describing or at least 
giving a reference to such an algorithm.

Tree rows -- real tree rows, i.e. a row of trees planted on purpose to 
function as a landscaping feature, not just some random trees which 
happen to be in a line -- are important landmarks and often show on maps 
as rows of green dots. However, the individual trees are typically too 
close to be shown at their real positions, so some generalization is 
required. Tagging the tree row provides such a generalization. I have no 
doubt that it is theoretically possible to synthesize tree-row objects 
from mapped trees, but I would guess that doing so with an acceptable 
number of false positives and negatives is close to a masters-thesis 
project.